performing-web-cache-poisoning-attack

High-risk offensive security skill. Its capabilities are aligned with its stated penetration-testing purpose and its tool references are mostly legitimate, but it equips an AI agent to perform real cache-poisoning and cache-deception attacks with potential impact on other users and exposure of authenticated content. Not malware, but dangerous by design and unsuitable for broad autonomous use.